Classifications

• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F9/00—Arrangements for program control, e.g. control units
  • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
  • G06F9/46—Multiprogramming arrangements
  • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
  • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
  • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
  • G06F9/505—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the load
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F9/00—Arrangements for program control, e.g. control units
  • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
  • G06F9/46—Multiprogramming arrangements
  • G06F9/461—Saving or restoring of program or task context
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F9/00—Arrangements for program control, e.g. control units
  • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
  • G06F9/46—Multiprogramming arrangements
  • G06F9/465—Distributed object oriented systems
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F9/00—Arrangements for program control, e.g. control units
  • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
  • G06F9/46—Multiprogramming arrangements
  • G06F9/466—Transaction processing
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F2209/00—Indexing scheme relating to G06F9/00
  • G06F2209/50—Indexing scheme relating to G06F9/50
  • G06F2209/5016—Session

Definitions

• the present invention relates generally to a method and apparatus for reconfiguring a server system comprising a plurality of servers for providing communication services.
• the server system is reconfigured dynamically such that shut-down periods for one or more servers can be avoided or at least be minimised.
• GPRS General Packet Radio Service
• CDMA Wideband Code Division Multiple Access
• GPRS General Packet Radio Service
• CDMA Wideband Code Division Multiple Access
• the data communicated in many new services may include voice, text, images, audio files and video files in various different formats and combinations.
• mobile instant messaging commonly known as "chatting”
• Presence services are rapidly becoming popular.
• Instant messaging is known from the world of fixed PCs (Personal Computers) , including message status reporting and various group and contact list features .
• Presence services involve information on the location of mobile terminals and enable users to receive messages according to their profile and availability.
• a user profile can be personal and may be defined by preferences, interests and hobbies, as well as more temporary factors, such as user availability and current moods .
• Messages and content services can also be delivered depending on the present location, availability and terminal capabilities. It can be readily understood that such services require the storage of considerable amounts of retrievable user-specific data, which in many cases need to be frequently updated due to their dynamic nature .
• server will be used hereafter to represent any hardware and/or software for storing and/or processing data.
• a server may be configured to provide one or more specific services.
• processing and storing load must be distributed over the servers . This is necessary in order to efficiently utilise available computing and storing resources, and to handle hotspots and avoid bottlenecks. As mentioned above, large amounts of data must be stored and should also be easy to find and retrieve.
• server systems must be reconfigured from time to time as the processing and storing load change, e.g., due to changing demands of service requests, added or removed subscribers and the introduction, modification or deletion of services.
• the workload on servers may increase rapidly so that individual servers are easily overloaded, at least for a short time, in particular popular web servers .
• overloading problems in servers basically two solutions are available. Firstly, an existing server may be upgraded to increase its computing and/or storing capabilities. However, the server will soon become overloaded again if the amount of service requests and/or needs for data storage continue to increase. Further upgrading is then required, which can be complex and costly to perform.
• a virtual server is a scalable server built on a cluster of real servers, which is transparent to end users such that the users see only a single virtual server.
• the front-end of the real servers is a node, sometimes referred to as a "load balancer" , configured to schedule service requests to the different real servers.
• Incoming service requests may involve processing tasks and storing tasks to be performed by the servers. Scalability can thus be achieved by transparently adding or removing servers in the cluster.
• it is a problem to efficiently distribute processing and storing tasks between a plurality of servers, yet enabling easy retrieval of stored data.
• Processing tasks may involve analysing service requests, processing of data and running certain applications for delivering requested services.
• “Storing tasks” may involve storing new client-specific, session-specific or configuring data, updating already stored data, and retrieving stored data. For example, a service request may require the retrieval of certain data which is used as input for executing a specific processing task or service application.
• Client data and session data is hereafter collectively referred to as "user data” .
• a server is often allocated to a client upon a login request .
• the allocation scheme used for selecting a server is normally based on the current load on a predetermined set of servers, such that the server having the lowest current load, with respect to memory resources and/or CPU (Central Processing Unit) capability, etc, is selected for a client or session.
• Server allocation is typically performed by using a central load manager node or the like. Stored user data must then be retrievable, i.e. it must be possible to' find the server in which the searched data was stored.
• Random Robin scheduling is only suitable for distributing processing load, since processing tasks are not affected by in which server they are performed.
• retrieving stored data in one of more servers cannot be done by using Round Robin but requires the use of specific pointers or references as described above.
• IP Internet Protocol
• a common basic problem with some of the other scheduling methods mentioned above is that they use IP (Internet Protocol) addressing for scheduling. Since a plurality of clients can reside behind a single IP address (proxy, NAT, etc.), these can neither be used for data distribution nor load sharing.
• Storing tasks' are therefore preferably scheduled by means of a hashing algorithm using some client or session identity as input .
• the hashing algorithm then provides a number identifying a selected server. If the same hashing algorithm is used again when retrieving data on a later occasion for the same client or session, the same server will be selected, provided that the servers have not been reconfigured since the searched data was stored. Thereby, the need for using specific pointers or references is eliminated.
• the hashing algorithm will most probably not provide the correct server identity anymore, if user data for a specific client or session was stored before the reconfiguration was made.
• a current solution for avoiding this problem is to shut down the server system from service requests for modifying the hashing algorithm and moving stored user data between servers, such that server selection becomes correct . This is a quite cumbersome and time consuming operation, which is undesirable since service requests cannot be attended meanwhile, resulting in lost revenue for the service provider.
• the reconfiguration operation is relatively complex involving a considerable risk for errors .
• the object of the present invention is to reduce or eliminate the problems outlined above, and to provide dynamic and efficient reconfiguration of a server system during simultaneous handling of service requests . Thereby, shut-down periods for One or more servers can be avoided, or at least be minimised.
• the server system comprises a plurality of servers being capable of performing at least one common storing task, and a scheduling unit configured to select servers for handling incoming service requests .
• a new server configuration is applied and a previous server configuration is saved.
• a server is selected 'for handling the received request, by using a first scheduling algorithm according to the new configuration. If the selected server was incorrect, a server is selected in a second attempt by using a second scheduling algorithm according to the saved previous configuration. If the correct server was selected by the second scheduling algorithm, user data associated with the received service request is moved from the selected server to the server selected by the first scheduling algorithm. If the selected server was incorrect also in the second attempt, a server can be further selected by using a scheduling algorithm according to a saved even earlier configuration.
• the scheduling algorithms are hashing algorithms .
• Using the hashing algorithms may include deriving a hash number from a user ID and calculating a server ID number from the derived hash number.
• the following algorithm may the be used:
• server ID hash (user ID) /modulo n (1)
• n is the number bf possible servers, and the modulo operator providing an integer between 0 and n-1.
• the new configuration may involve a changed number of servers, and in the second scheduling algorithm, n • is then changed accordingly from the first scheduling algorithm.
• the service requests may involve storing tasks such as storing new data, updating already stored data, and retrieving stored data.
• Applying a new server configuration may further include storing new configuration data in a central administrator or the like, which can be accessed by the servers.
• Each server may comprise a local administrator or the like, being adapted to receive new configuration data from the central administrator.
• New configuration data can be pushed from the central administrator to the servers, or be retrieved by sending a request from a server.
• the central administrator may comprise means for pushing new configuration data to one or more servers.
• the local administrator may comprise means for sending a request for new configuration data to the central administrator.
• the inventive procedure may further be executed by means of a computer program comprising a software code being adapted to perform the method in a server system.
• the present invention is capable of simultaneously handling a dual configuration of a server system.
• the system is to be reconfigured by adding and/or removing one or more servers, the previous old configuration is saved. Thereby, requested data can be retrieved by using the old configuration, if the requested data could not be found by using the new configuration.
• Stored data is gradually and dynamically reorganised from the old* configuration to the new one with a minimum of impact on the system operation.
• a better distribution of data may also be achieved when reconfiguring a server system, since data associated with new users or sessions can be efficiently distributed on existing servers as well as on any new added servers .
• Fig. 1 is a schematic overview of a communication scenario in which the present invention can be used.
• Fig. 2 is a block diagram of an exemplary server system according to one embodiment .
• Fig. 3 is a flow chart illustrating a procedure for handling a service request .
• Fig. 4 is a block diagram illustrating centralised administration of configuration data.
• FIG. 1 a schematic communication scenario is illustrated in which the present invention can be used.
• a plurality of client terminals 100 are connected to a backbone network 102, such as the Internet.
• the client terminals 100 may be fixed or mobile, such as wired terminals, or wireless terminals connected to a mobile access network 104 over an air interface, as indicated in the figure.
• a plurality of service providers 106 are likewise connected to the backbone network 102, each comprising one or more 1 servers for executing telecommunication services requested for the client terminals' 100.
• various further networks and/or nodes are typically involved in communication routes between clients and service providers, although not shown here for the sake of simplicity.
• a client terminal 100 may initiate a specific telecommunication service by sending a service request over the backbone network 102 to a service provider 106.
• the contacted service provider then activates one or more suitable service applications in response thereto.
• Activating a service application may involve various processing and storing tasks, as performed in one or more servers.
• a service application may also be triggered without a preceding terminal request, such as through a "push" mechanism as used for example in the context of WAP (Wireless Application Protocol) .
• a service request may thus originate from another service provider or network operator, e.g. needing user data of a certain client or ongoing session.
• a service provider may want to transmit certain information to mobile stations as they enter a specific area. In that case, the service provider will request for needed user data, such as terminal capabilities and client profile, e.g., including predefined preferences and interests .
• service providers are typically equipped with a plurality of servers in order to handle service requests from clients and other service providers.
• the same functionality is duplicated in several servers, thereby being capable of performing the same or similar service tasks simultaneously for plural clients, be it processing or storing tasks.
• the present invention is concerned* with a method and apparatus for enabling efficient reconfiguration of plural servers such that any shut-down periods can be avoided or at least be minimised.
• a simplified server system 200 is , illustrated for providing one or more telecommunication services for client terminals.
• the server system 200 comprises a scheduling unit 202 for scheduling incoming service requests involving one or more • storing tasks .
• the scheduling unit 202 is connected to a set of servers 204 being capable of performing at least one common storing task.
• the servers 204 may in addition be capable of performing other tasks or operations as well, which is however not within the scope of the present invention.
• the servers 204 may further be designated for different service applications, and may comprise Session servers, Instant Messaging servers, Presence servers, etc.
• the server system 200 may also include further components such as access nodes and different servers designated for different request handling operations . However, the functionality of these further components are generally outside the scope of the present invention, which are therefore not described here .
• the scheduling unit 202 may reside in an access node of a service provider, receiving service requests from clients, service providers or network operators.
• the scheduling unit 202 may also reside in a processing server being configured to forward certain service requests to other servers for further treatment .
• the scheduling unit 202 is configured to use scheduling algorithms for .selecting specific servers 204 for incoming service requests, using some predefined client or session identity as input.
• a scheduling algorithm then provides a number identifying a selected server.
• the scheduling algorithm provides that when stored data for a client or session is retrieved, the correct server will be selected, if the number of servers 204 and their identities are the same as when the searched data'was stored. Further, if a storing task involves storing new user data, the scheduling algorithm will provide that the new data is stored in the same server as old user data was stored, thereby collecting all user data for a specific client or session in the same server.
• a hashing algorithm means that a hash number is derived from a predetermined identity code of a client or a session, which will hereafter be collectively referred to as a "user ID" .
• a server identity can be determined by calculating a corresponding server ID number by means of a predetermined algorithm or formula based on the derived hash number.
• the hash number can be a simple checksum or the like for a binary code of the user ID.
• a server 204 is determined to perform a storing task from the following algorithm:
• server ID hash (user ID) modulo n (1)
• n is the number of possible servers in the set .
• the server system needs to be reconfigured by changing the number of servers, the used hashing algorithm must also be modified in order to distribute storing load between the new number of servers.
• n must be changed according to the new number of servers. Thereby, inputting a user ID in the modified hashing algorithm will most probably provide a different server ID than before .
• the present invention provides a dynamic solution allowing simultaneous operation of handling service requests, 'thereby avoiding unwanted shutdown periods.
• the hashing algorithm is modified according to the new configuration, and the old configuration is saved in a memory 206 in the scheduling unit 202. Saving the old configuration includes saving at least the old hashing algorithm and its parameters, and possibly also information on the server system structure of the old configuration.
• the scheduling unit 202 When receiving a service request involving a storing task for a client or session as identified by a user ID, the scheduling unit 202 makes a first attempt to find the correct server by using the modified new hashing algorithm. If the expected server is not found, i.e. the server where searched data or old user data was stored, the old hashing algorithm according to the saved old configuration is used in a second attempt. When the correct server is found and selected, for the storing task, data associated with the concerned client or session is moved from that server to a new server corresponding to the new hashing algorithm. If the new hashing algorithm happens to select the same server as the old hashing algorithm, the data will of course remain there.
• a service request is received in a scheduling unit 202.
• a storing task must be executed for the service request in one of the servers 204, such as retrieving already stored user data.
• a first attempt is made to find the correct server where the searched data is stored.
• the new hashing algorithm is then applied to select a server for the received service request, using a corresponding user ID as input .
• the user ID can normally be extracted from the received service request .
• step 304 it is determined whether the correct server was found and selected in step 302 by the new hashing algorithm. If so, the storing task can be executed accordingly in- a step 306, after which the procedure may return to the first step 300 for receiving a next service request. If it is determined in step 304 that the selected server was incorrect, a second attempt is made to find the correct server in a step 308, by applying a saved previous hashing algorithm on the received service request .
• step 310 It is then determined in a next step 310, whether the correct server was found and selected in step 308 by the previous hashing algorithm. If so, user data associated with the service request is moved, in a step 312, from the old server to the new server selected in step 302 by the new hashing algorithm.
• the moved data may embrace all user data of a client or session for which the service request was directed, even if the 'storing task to be performed is concerned with only some of that data.
• the storing task is also executed accordingly in a step 314, after which the procedure may return to the first step 300 for receiving a next service request.
• steps 312 and 314 can be performed in reverse order, if suitable. For example, it may be efficient to first retrieve searched data in a storing task before moving the data to another server. If it was determined in step 310 that the correct server was still not selected in step 308, it may be determined in a step 316 whether a new attempt should be made to find the correct server. If so, the procedure returns to step 308 for applying a further previous hashing algorithm, which may be an even older saved hashing algorithm, as explained above. If determined in step 316 that no new attempt is to be made, e.g. due to time-out or count-out, the storing task cannot be performed for the received service request . A new request may then be received by returning to the first step 300.
• reconfiguring a server system may involve changes in offered services. For example, new services may be added and present services may be modified or omitted. In a system with plural servers, such reconfiguration requires that all servers are updated with the new configuration. In previous solutions, relevant new configuration data is stored in all servers. In the present invention, all configuration data may instead be stored in a central administrator or the like, which can be accessed from the servers .
• FIG. 4 a schematic block diagram illustrates a centralised administration of configuration data.
• a group of servers 400 is shown, 'each being capable of handling at least one common service request, such as the servers 204 in Fig. 2.
• Each server 400 comprises a local administrator 402 which is connected to a central administrator 404 and having functionality for accessing configuration data there from.
• the central administrator 404 may be a separate node in the server system or may reside in an access node or in a server, e.g. together with a scheduling unit 202 as described above.
• the servers 400 may need access to certain configuration data in order to invoke requested services. However, no such configuration data is stored locally in the servers according to one embodiment . Instead, all configuration data being valid for the servers 400 is stored in the central ' administrator 404.
• any configuration data required for invoking a requested service in a server 400 can be retrieved from the central administrator 404 by sending a request for the configuration data from its local administrator 402. Only a reference to the central administrator 404 is needed in the local administrators 402. Configuration data can also be "pushed" to one or more local administrators 402 from the central administrator 404, i.e. without a preceding request from the local administrators 402.
• configuration data is efficiently ' stored in only one place instead of being duplicated in the servers 400.
• Any updating of configuration data can easily be done in the central administrator 404, which will have an accurate overview of all current services and servers . Errors and conflicts between servers having different configuration versions can thereby be avoided.
• the need for synchronous updates in plural servers is eliminated.
• the present invention may be used to great advantage in telecommunication services defined in the context of "Wireless Village Server" , such as those relating to Instant Messaging, Presence Information and Shared Content .
• the present invention is also applicable to any scalable systems for data distribution.
• servers can thus be dynamically added and/or removed without shutting down the server system, at least for extensive time periods. In this way, a high level of service can be maintained.
• the present invention can advantageously be used in a server system for distributing data storage and processing load between plural servers which is described in Applicant's co-pending PCT application .
• a primary server for performing a processing task is assigned using a first scheduling algorithm, which is capable of selecting any primary server.
• a secondary server for performing a storing task is assigned using a second scheduling algorithm, which is capable of selecting a specific secondary server corresponding to a client involved in that storing task.

Landscapes

• Engineering & Computer Science (AREA)
• Software Systems (AREA)
• Theoretical Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Engineering & Computer Science (AREA)
• General Physics & Mathematics (AREA)
• Computer And Data Communications (AREA)
• Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
• Hardware Redundancy (AREA)
• Multi Processors (AREA)
• Small-Scale Networks (AREA)
• Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)

Priority Applications (4)

Applications Claiming Priority (2)

Publications (1)

Family

ID=20286945

Family Applications (1)

Country Status (7)

Cited By (2)

Families Citing this family (14)

Citations (1)

Family Cites Families (15)

• 2002
  • 2002-02-13 SE SE0200417A patent/SE0200417D0/xx unknown
  • 2002-06-27 JP JP2003568530A patent/JP4098723B2/ja not_active Expired - Fee Related
  • 2002-06-27 GB GB0416215A patent/GB2400951B/en not_active Expired - Fee Related
  • 2002-06-27 WO PCT/SE2002/001279 patent/WO2003069473A1/en active IP Right Grant
  • 2002-06-27 US US10/504,086 patent/US7739364B2/en active Active
  • 2002-06-27 ES ES200450045A patent/ES2264627B2/es not_active Expired - Fee Related
  • 2002-06-27 AU AU2002345485A patent/AU2002345485A1/en not_active Abandoned

Patent Citations (1)

Non-Patent Citations (3)

Also Published As

Similar Documents

Legal Events